CN110320892B - Sewage treatment equipment fault diagnosis system and method based on L asso regression - Google Patents
Sewage treatment equipment fault diagnosis system and method based on L asso regression Download PDFInfo
- Publication number
- CN110320892B CN110320892B CN201910635096.1A CN201910635096A CN110320892B CN 110320892 B CN110320892 B CN 110320892B CN 201910635096 A CN201910635096 A CN 201910635096A CN 110320892 B CN110320892 B CN 110320892B
- Authority
- CN
- China
- Prior art keywords
- data
- sewage treatment
- equipment
- model
- regression
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B23/00—Testing or monitoring of control systems or parts thereof
- G05B23/02—Electric testing or monitoring
- G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
- G05B23/0259—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterized by the response to fault detection
- G05B23/0262—Confirmation of fault detection, e.g. extra checks to confirm that a failure has indeed occurred
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B2219/00—Program-control systems
- G05B2219/20—Pc systems
- G05B2219/24—Pc safety
- G05B2219/24065—Real time diagnostics
Abstract
The invention belongs to the field of equipment fault diagnosis, and particularly relates to a fault diagnosis system and method of sewage treatment equipment based on L asso regression, wherein the fault diagnosis system comprises a sewage sub-treatment system, a plurality of supervision clients and a plurality of data acquisition devices, the sewage treatment subsystem comprises a central console, a communication server, a total data storage server and a data acquisition server, the central console comprises a control module, a data interface module, a data classification module, a machine learning module and a data communication module, and the machine learning module performs learning and training on the data acquisition devices by acquiring data on the data acquisition devices, so that a plurality of sub-regression prediction models are constructed, and fault types of various sewage treatment equipment are predicted.
Description
Technical Field
The invention relates to the field of equipment fault diagnosis, in particular to a sewage treatment equipment fault diagnosis system and method based on L asso regression.
Background
The water environment treatment is always an important project for improving the livelihood and building the ecological civilization of China, wherein the sewage treatment is the first important task. At the present stage, with the gradual improvement of the social level and the deepening of the industrial development degree of China, the social water demand is large, the industrial sewage discharge amount is increased, so that the water resource environment of China is damaged to a certain degree, the problems of water resource shortage, water ecological damage, water environment pollution and the like are more and more prominent, and the water safety guarantee faces a serious challenge. In government water treatment planning, the treatment of black and odorous water is the most important part of the current urban water treatment work. Therefore, in order to promote the construction of ecological civilization in China and reduce the waste of water resources, sewage treatment work needs to be done, the cyclic application of water resources is enhanced, and the sustainable development planning in China is achieved.
In recent years, with the progress of water treatment work, the large-scale construction of infrastructures such as a plurality of intelligent water plants and the like, the condition of sewage treatment is greatly improved. However, the sewage treatment process is a dynamic nonlinear reaction process with various variables and time-lag characteristics, and various faults are easy to occur in the actual operation process. For example, the water supply network has abnormal events, and the water quality is abnormal due to external immersion, aging of pipeline facilities and water body retention, so that the equipment is further lost; the water pressure is abnormal due to sudden unconventional water use and unreasonable pressure regulation, so that the pipe explosion and blockage of the equipment are caused; the phenomena of overcurrent equipment faults and the like caused by the improper valve action, the blockage of exhaust equipment, the blockage of a pressure reducing valve and the like. The frequent occurrence of the faults can influence the long-term stable operation of the sewage treatment plant, and if the equipment faults cannot be found in time and effective countermeasures can be taken, the process of sewage treatment can be stopped, and even the sewage with unqualified discharged water quality can pass through. Therefore, in the process of treating water quality by sewage equipment, the running state of the equipment is diagnosed in real time, the fault of the equipment can be accurately judged and diagnosed in time, the equipment is ensured to be treated in time after the fault occurs, and the method has important significance for the long-term stable operation of a sewage treatment plant
In the existing sewage treatment field, the treatment for faults still relies heavily on manual treatment means, and the defects of efficiency and quality exist, such as: (1) the faulty device is not handled in time. In order to examine and repair whether the equipment normally operates, the current traditional water business enterprise method relies on manually recording equipment fault data one by one, or after the equipment transmits the data, workers arrange the obtained report, and periodically patrol and maintain the equipment in stages according to a processing flow, so that the workload is large and the efficiency is low. Due to the continuous improvement and the continuous increase of the quantity and the scale of the current water business enterprise machine equipment, manual treatment is difficult, equipment failure can be detected after a certain time, and the treatment process is influenced to a certain extent. (2) The faulty device is not in place. Under the condition that equipment instruments of a current sewage treatment plant are numerous, data volume is large, and equipment structure is complex, most of existing equipment detection means still rely on manual experience, mainly include online monitoring and judgment, visual, olfactory and tactile perception experiences of inspection personnel, equipment fault indication and the like, and lack of means for accurate positioning for faulty equipment, so negligence and errors are easy to occur, if the equipment or type of fault is judged incorrectly, maintenance time is prolonged or spare parts are wasted, even new faults are caused, and maintenance difficulty is increased.
Disclosure of Invention
In view of the above, the invention provides a fault diagnosis system and method for sewage treatment equipment based on L asso regression, the fault diagnosis system can predict the operation condition of equipment in a normal state according to the current sewage treatment condition, diagnose in combination with the real-time equipment operation condition according to the prediction result, and can timely and accurately judge and diagnose the occurrence of equipment faults in advance, and meanwhile, based on the deployment of the system, the system automatically collects data of the equipment and the production environment, reduces the hysteresis of equipment maintenance detection, and can timely issue fault early warning information according to the judgment result of the method, so that maintenance personnel can timely and accurately operate and maintain, and can timely process the equipment after the equipment faults occur, thereby avoiding accidents.
The fault diagnosis system comprises a sewage treatment subsystem, a plurality of supervision clients and a plurality of data acquisition devices arranged in sewage treatment equipment and a sewage treatment pool in a sewage plant; wherein the data acquisition device is respectively positioned at a plurality of monitoring points. The monitoring points comprise: a grit chamber, a disinfection tank, a clarification tank, a filter tank and other treatment tanks, a sludge pump room, a grid pump room and other plants and a remote data acquisition device.
Further, the sewage treatment subsystem comprises a central console, a communication server, a total data storage server and a data acquisition server;
furthermore, the central console comprises a control module, a data interface module, a data classification module, a machine learning module and a data communication module;
the central control console is a large computer cluster, is a control center of the sewage treatment subsystem, is responsible for controlling and coordinating the work flow and data flow among the modules, transmits data with each server through the concentrator, and issues real-time information or early warning information.
Furthermore, the control module is used for controlling and coordinating the workflow and data flow among the modules, carrying out data transmission with each server through the concentrator, and issuing real-time information or early warning information of the sewage treatment equipment according to the collected data of each server;
further, the data interface module is used for receiving a data collection instruction of the control module and accessing the total data storage server according to the instruction to obtain corresponding data;
further, the data classification module is used for classifying the received sewage equipment data and sewage treatment environment data and taking the classified data as input data of the machine learning module;
specifically, the device data processing method is responsible for classifying received device data and environment data, and the device data is subjected to data preprocessing according to the device type to which the device data belongs, so that a feature vector set is obtained and is used as input data of the machine learning module.
Further, the machine learning module is used for constructing a sewage treatment equipment fault prediction model based on L asso linear regression and predicting sewage treatment equipment fault information;
specifically, the method is responsible for constructing a fault prediction model, training historical data on the model to obtain an available prediction model, and inputting real-time data serving as the model in practical application to obtain a prediction result.
Further, the data communication module is used for receiving the equipment failure prediction information of the machine learning module and the position information of the sewage treatment equipment and the real-time data of the sewage treatment equipment and the sewage treatment environment acquired by the data interface module, and packaging and transmitting the information to the communication server;
further, the total data storage server is used for storing historical water quality data, equipment running state data and sewage treatment subsystem running data and sending the historical water quality data, the equipment running state data and the sewage treatment subsystem running data to the central console through the data interface module; the data can be transmitted to the platform for processing in real time.
Further, the communication server is responsible for data communication between a data communication module of the central console and each supervision client; the monitoring point checking instruction of the client supervisor can be received, and the prediction information and the monitoring information of the central console can be returned in real time.
Furthermore, the data acquisition server is used for acquiring real-time data of the sewage treatment equipment and the sewage treatment environment acquired by the data acquisition devices.
The monitoring system is used for collecting real-time data of sewage treatment equipment and a sewage treatment environment from a production environment, wherein the equipment data comprise pressure indexes of pressure regulating equipment, instantaneous flow of overflowing equipment, equipment running time, equipment running states and the like, the water quality data comprise nitrogen and phosphorus content, mineral content, oxygen consumption, sludge deposition and the like, the monitoring system is specifically interconnected with remote monitoring terminals of sewage plant equipment and a sewage treatment pool at monitoring points, and data of instrument equipment at the monitoring points are collected and uploaded to a total data storage server.
Furthermore, the step of issuing real-time information or early warning information according to the acquired data of each server comprises the step of packaging the fault information of the sewage treatment equipment predicted by the machine learning module, wherein the fault information comprises a prediction result, a model identifier and time sequence information, and the prediction result, the model identifier and the time sequence information are transmitted to the central control platform; the central control platform obtains corresponding equipment information according to the predicted fault information of the sewage treatment equipment, the equipment information comprises the position of a monitoring point where the equipment is located, current state data and environmental data of the equipment, the data and the predicted information are packaged into early warning information, and the early warning information is transmitted to the data communication module; the data communication module transmits the early warning information to a communication server; and the communication server sends the early warning information to the supervisor client.
Further, the machine learning module includes a plurality of sub-modules,
the model construction submodule is used for constructing a sewage treatment equipment fault prediction model based on L asso linear regression;
a model training submodule: training a fault prediction model of the sewage treatment equipment by using historical data of the sewage treatment equipment and the sewage treatment environment as training set data, and acquiring parameters of the fault prediction model;
the model prediction submodule predicts the running time of the equipment by utilizing a trained sewage treatment equipment fault prediction model based on L asso linear regression;
a run-time residual calculation submodule: calculating residual errors of historical fault equipment running time and normal equipment running time to obtain a residual error sequence and residual error absolute values of actual running time and predicted running time of real-time computing equipment;
a threshold construction submodule: calculating a first four quantile point Q1, a median Q2 and a third four quantile point Q3 of the residual sequence, wherein the IQR is 1.5 (Q3-Q1), the Q1-IQR is a lower inner limit, and the lower inner limit is a fault threshold;
and a fault information judgment submodule: and comparing the residual absolute value with a fault threshold, and outputting fault early warning information if the residual absolute value is greater than the fault threshold.
The invention also provides a fault diagnosis method of sewage treatment equipment based on L asso regression, which comprises the following steps:
a. acquiring a large amount of data of sewage treatment equipment and sewage treatment environment from a plurality of data acquisition devices of the sewage treatment equipment and a sewage treatment pool in a sewage plant, wherein the data comprises the data of the sewage treatment equipment and the data of the sewage treatment environment;
b. performing data preprocessing on the equipment data according to the type of the equipment, namely standardizing the characteristics of each dimension of the data according to the distribution condition of the characteristics of the type of the equipment to obtain a characteristic vector set as training set data;
c. constructing a fault prediction model of the sewage treatment equipment based on L asso linear regression, training the model by using training set data, and acquiring regression parameters of the model;
d. acquiring real-time data of the sewage treatment equipment and the sewage treatment environment from a plurality of data acquisition devices of the sewage treatment equipment and the sewage treatment pool in a sewage plant, performing a data preprocessing process consistent with the step b, and predicting the operation time of the equipment by taking the obtained feature vector set as the input of a trained model;
e. and calculating the residual absolute value of the actual running time and the predicted running time of the equipment, comparing the residual absolute value with a fault threshold, and outputting fault early warning information if the residual absolute value is greater than the fault threshold.
Further, the data acquisition part in the steps a and d is completed by a data acquisition server, and the data used in the steps are taken from a total data storage server.
Further, the data preprocessing process in the step b is to standardize the features of each dimension of the data according to the distribution condition of the features of the data types;
further, the L asso linear regression-based sewage treatment plant fault prediction model in the step c comprises:
wherein, YiResponse variable for the ith test sample, specifically predicted working time, X, of the feature vector concentration deviceijThe ith test sample corresponding to the jth second regression parameter is specifically real-time data of equipment and environment in the data feature vector;
estimated values of the first regression parameters representing the lasso linear regression model βjAn estimate of a jth second regression parameter representing a lasso linear regression model;iis the error variable for the ith training sample,i∈N(0,σ2) (ii) a i represents the serial number of a test sample, namely the serial number of real-time data of equipment and environment in the data characteristic vector to be predicted, and j represents the serial number of a model regression parameter; p represents the number of model regression parameters.
Further, the obtaining manner of the first regression parameter and each second regression parameter estimation value in step c includes:
wherein, yiThe response variable of the ith training sample is specifically the working time of equipment in the data characteristic vector in the training process; x is the number ofijα is the ith training sample corresponding to the jth second regression parameter, specifically the real-time data of the equipment and environment in the data feature vectoriβ first regression parameter of lasso linear regression model for ith training samplejThe j-th second regression parameter in the lasso linear regression model is represented, t is a constraint parameter, i represents the serial number of the training sample, and j represents the serial number of the model regression parameter; t denotes a constraint parameter.
Further, the constraint parameter t in the parameter estimation formula is calculated by using a rolling cross validation method, and the establishment process comprises:
dividing the training data with the number T into three parts of equal size or approximately equal size T1, T2 and T3;
t1 partial data is selected as constraint parameter T with initial value
βiFor the model regression parameters obtained by the least square method, decreasing the constraint parameter t by a certain step length, and training the new model as the constraint parameter of the new model to obtain evaluation models with different parameters t; wherein, the step length can be set according to actual needs.
The T2 partial data is used for predicting and evaluating the evaluation model obtained by the T1 partial data, and prediction is carried out once when a sample observation value is sequentially added, so that the square sum SE of prediction errors of the T1 evaluation models is obtained;
the T3 partial data is used for obtaining final parameters, the parameters of the evaluation model with the lowest SE and the parameters weighted and averaged according to the SE are taken to train the training data consisting of T1 and T2, the obtained model is used for prediction evaluation by using T3, and the parameters with the lowest SE are the final parameters.
Further, a lasso loss function is used as an objective function, a prediction model is trained by using a large amount of data to solve a solution which meets the minimum value of the objective function, an estimated value of a first regression parameter of the prediction model and an estimated value of each second regression parameter are obtained, so that a regression parameter set which meets the lasso constraint is obtained, and the loss function is expressed as:
β in the above equation is a second regression parameter vector of length n that does not include coefficients of the intercept term, and θ is an input vector of length n +1 that includes coefficients of the intercept term θ0M is the number of training samples, n is the number of features, | | | β | | purple1L representing parameter β1Norm, λ, is the constraint term of the input vector.
Further, the solution satisfying the minimum value of the loss function is obtained by a coordinate descent method.
Further, the obtaining process of the failure threshold value in step e includes:
calculating residual absolute values of historical fault equipment running time and normal equipment running time to obtain a residual absolute value sequence;
establishing a box-type graph statistical model, calculating a first four-quantile point Q1, a median Q2 and a third four-quantile point Q3 of a residual absolute value sequence, and making the IQR (iQR) 1.5 (Q3-Q1) and Q1-IQR as a lower inner limit; the lower inner limit is taken as the failure threshold.
The invention has the beneficial effects that: the invention connects different equipment in the sewage treatment environment with the sewage treatment subsystem, automatically acquires the data of the equipment and the production environment, accords with the characteristics that the water business industry generates a large amount of real-time data in actual business and the production environment changes rapidly, reduces the hysteresis of equipment maintenance and detection, relieves the problem of low manual treatment efficiency, can pertinently predict the operation condition of the equipment, diagnoses the operation condition of the equipment by combining the real-time data, can accurately judge and diagnose the occurrence of equipment failure in advance, issues failure early warning information in time, effectively meets the requirements of real-time monitoring and diagnosis of the fault problem of the sewage treatment equipment in the sewage treatment process, provides diagnosis help with a certain foundation for the engineering management and treatment decision of sewage treatment, and assists the adjustment and maintenance work of instrument equipment in the real-time environment to a certain extent, the threat of the sudden problem to the stability of the sewage treatment project is reduced.
Drawings
The invention is further described below with reference to the following figures and examples:
FIG. 1 is a schematic diagram of the system architecture of the present invention.
FIG. 2 is a schematic flow diagram of the system of the present invention.
Detailed Description
The invention is further described below with reference to the following figures and examples:
as shown in FIG. 1, the invention provides a fault diagnosis system of sewage treatment equipment based on L asso regression, which comprises a sewage treatment subsystem, a plurality of supervision clients and a plurality of data acquisition devices arranged in the sewage treatment equipment and a sewage treatment pool in a sewage plant, wherein the sewage treatment subsystem comprises a central console, a communication server, a total data storage server and a data acquisition server.
Clients 1-n in the figure, and devices 1-m; the specific number is not limited, but is only used for indicating that the number of the clients is multiple, the number of the clients and the devices is not necessarily related, and the devices in the figures comprise sewage treatment devices and data acquisition devices.
The central console comprises a control module, a data interface module, a data classification module, a machine learning module and a data communication module; and the central console acquires data of different equipment and the environments of the warehouses in batches from the total data storage according to a cyclic access method and a certain time interval. On the other hand, the central console acquires corresponding equipment information according to the fault information, wherein the equipment information comprises the position of the monitoring point where the equipment is located, current state data and environment data of the equipment, the data and the prediction information are packaged into early warning information, and the early warning information is transmitted to the data communication module.
The control module is used for controlling and coordinating the work flow and data flow among the modules, carrying out data transmission with each server through the concentrator, and issuing real-time information or early warning information of the sewage treatment equipment according to the collected data of each server;
the data interface module is used for receiving a data collection instruction of the control module and accessing a total data storage server according to the instruction to obtain corresponding data; and after receiving the data, the data interface module transmits the data to the data classification module, classifies the data according to the source of the sewage treatment equipment in the data, transmits the packed data blocks, standardizes the data to obtain a feature vector set of the data, and uses the feature vector set as input data in the machine learning module.
The data classification module is used for classifying the received sewage equipment data and sewage treatment environment data and taking the classified data as input data of the machine learning module;
the machine learning module is used for constructing a sewage treatment equipment fault prediction model based on L asso linear regression and predicting sewage treatment equipment fault information, the machine learning module is used for distributing the received feature vector set to the corresponding regression prediction model in the model set, the model predicts the current state of the equipment in real time according to the feature vector, and compares the predicted running time according to the threshold value of the equipment.
The data communication module is used for receiving equipment failure prediction information of the machine learning module and position information of the sewage treatment equipment and real-time data of the sewage treatment equipment and the sewage treatment environment acquired by the data interface module, and packaging and transmitting the information to the communication server;
the total data storage server is used for storing historical water quality data, equipment running state data and sewage treatment subsystem running data and sending the historical water quality data, the equipment running state data and the sewage treatment subsystem running data to the central console through the data interface module;
the communication server is responsible for data communication between a data communication module of the central control console and each supervision client;
the data acquisition server is used for acquiring real-time data of the sewage treatment equipment and the sewage treatment environment acquired by the data acquisition devices.
The machine learning module includes a plurality of sub-modules,
the model construction submodule is used for constructing a sewage treatment equipment fault prediction model based on L asso linear regression;
a model training submodule: training a fault prediction model of the sewage treatment equipment by using historical data of the sewage treatment equipment and the sewage treatment environment as training set data, and acquiring parameters of the fault prediction model;
the model prediction submodule predicts the running time of the equipment by utilizing a trained sewage treatment equipment fault prediction model based on L asso linear regression;
a run-time residual calculation submodule: calculating residual errors of historical fault equipment running time and normal equipment running time to obtain a residual error sequence and residual error absolute values of actual running time and predicted running time of real-time computing equipment;
a threshold construction submodule for calculating a first four-quantile point Q1, a median Q2 and a third four-quantile point Q3 of the residual sequence, wherein the IQR is 1.5 × (Q3-Q1), the Q1-IQR is taken as a lower inner limit, and the lower inner limit is taken as a fault threshold;
and a fault information judgment submodule: and comparing the residual absolute value with a fault threshold, and outputting fault early warning information if the residual absolute value is greater than the fault threshold.
In this embodiment, the process of establishing the fault threshold by the threshold establishing sub-module using the box line graph model is as follows:
calculating residual absolute values of historical fault equipment running time and normal equipment running time to obtain a residual absolute value sequence, and sequencing the sequence according to the size;
establishing a box-type graph statistical model, calculating a first four-quantile point Q1 of a residual absolute value sequence, wherein the positions of the first four-quantile point Q1 are (n +1)/4 th positions in the sequence, a median Q2 is 2 (n +1)/4 th positions in the sequence, a third four-quantile point Q3 is 3 (n +1)/4 th positions in the sequence, and the IQR is made to be 1.5 (Q3-Q1), so that Q1-IQR is a lower inner limit, and Q3+ IQR is an upper inner limit; the equation n is the number of data in the sequence.
Taking down the inner limit as a fault threshold;
normally, the values outside the upper and lower limits are all abnormal values in the data, but since it is only necessary to determine whether the working time residual is within the fault range, that is, whether the working time residual is the lowest normal value in the fault data sequence, it is only necessary to determine whether the working time residual is the deviation exceeding a certain degree, and therefore, the fault time residual exceeding the lower limit can be used as the basis for the determination.
In conclusion, in the machine learning module, the fault threshold is established based on the historical data of the equipment fault, and the model construction sub-module completes construction of each equipment prediction model, so that the running condition of the equipment in the sewage treatment process can be monitored in real time through system deployment, and early warning is performed on the equipment which is likely to have the fault.
As shown in fig. 2, an implementation is provided, where a data acquisition server acquires real-time data of each sewage treatment device from each data acquisition terminal, a data classification module classifies the real-time data according to the source of the device, packages the classified data into data blocks, sends the data blocks to a fault diagnosis model of a machine learning module, performs data normalization on the data blocks by the fault diagnosis model, predicts the current state of the device, and returns a prediction result, so that a client can check the current state of a certain device.
The invention also provides a fault diagnosis method of the sewage treatment equipment based on L asso regression, which comprises the following steps:
a. a large amount of data of sewage treatment equipment and sewage treatment environment are collected from the production environment, wherein the data comprise equipment data and water quality data, the equipment data comprise pressure indexes of pressure regulating equipment, instantaneous flow of overflowing equipment, equipment running time, equipment running states and the like, and the water quality data comprise nitrogen and phosphorus content, mineral content, oxygen consumption, sludge deposition and the like.
In this embodiment, the data acquisition server in the sewage treatment subsystem is connected with the remote monitoring terminal of each monitoring point in the production environment, the on-the-spot equipment of sensor real-time monitoring at terminal, including blower equipment, pressure regulating equipment, overflow equipment, water quality testing equipment, activation testing equipment etc. gather in real time and receive the data in the equipment instrument, such as pressure index of pressure regulating equipment, overflow equipment's instantaneous flow, equipment operating duration, equipment running state etc. water quality data is nitrogen phosphorus content, mineral content, oxygen consumption, sludge deposition etc. realize multisource data collection, in time reflect the condition of real-time industrial environment.
After the equipment data and the environmental data are transmitted to the data acquisition server in real time, the data acquisition server is connected with the total data storage server, the data stored in the cache of the server are sent to the total data storage server in time, a database in the total data storage server classifies and stores the data according to the specific equipment of the data source and the data acquisition time, and an instruction for accessing the data by the central control platform is waited.
Further, after the acquisition of the required data is completed,
b. performing data preprocessing on the equipment data according to the type of the equipment to which the equipment data belongs to obtain a feature vector set serving as training data;
in this embodiment, the central control platform acquires the sewage treatment data from the total data storage server through the data interface module, and the sewage treatment data can be divided into historical data and real-time data from the time perspective, wherein the historical data is used for a model building part of the machine learning module, and the real-time data is used for a model prediction part. The data interface module transmits the data to the data classification module, and the data classification module is used for distinguishing and classifying different data, and carrying out targeted preprocessing on the data to obtain a characteristic vector of the data so as to carry out further processing.
b01. Firstly, according to the equipment and the environment source of the Data, different labels are marked on the Data, the specific label content is distinguished according to the equipment or the environment source of the Data, the specific label content comprises the equipment number or the sewage treatment plant number, and then the Data is packaged into different types of Data blocks which can be expressed as DataBloc { Data, T }, wherein Dat is original Data, T is a Data type, and each Data type corresponds to a specific equipment.
b02. And performing data preprocessing on the classified data blocks to obtain a feature vector set.
In a blower, for example, the phase current of a motor is in units of A, usually more than 200, and the oil pressure of a gear pump is in units of kilograms per square centimeter, usually one decimal place later. The two index variables in the equipment operation environment have large dimensional difference, and when the degree of reflection of the index on the equipment operation condition is measured, the variable with large dimension may have more obvious effect under certain consideration, but the effect may be only a numerical relationship rather than an actual effect relationship. It is therefore necessary to pre-process the data before it is used.
In the embodiment, a standardization method is adopted to standardize the data, the standardization is essentially linear transformation, and for a specific sample data, the range, the maximum value, the mean value and the variance of the sample data are all fixed, so that the sample data can be regarded as one-time scaling when the standardization is carried out, the sequencing of the original data is not changed, the linear relation is not changed, the interaction relation between the data is not changed, and the precision and the operation speed of a prediction model are improved.
The normalization process normalizes the features of the data, specifically, normalizes the features of each dimension in the data, and at this time, the data can be expressed as:
where n is the number of samples in the sample,
where i is the ith data item and j is the jth feature in the data.
Standardizing all data by using a standardization method, specifically:
wherein
A jth feature representing the normalized ith data item,
a jth feature representing an unnormalized ith data item,
represents the mean of the feature.
After the data in the data block is preprocessed, a feature vector set of the data is obtained, and a new data block packed by the feature vector set can be represented as Databloc { X, T }, where X is a feature vector set matrix and T is a data type of the data block. And the data classification module sends the processed data blocks to the machine learning module as a training set.
c. Establishing an equipment prediction model for the sewage treatment equipment based on a lasso linear regression model, and acquiring model parameters by using training set data;
under normal conditions of equipment, the running time generally requires to be controlled within a range, random fluctuation occurs under the influence of an external environment and certain degradation conditions, and a large number of extreme anomalies are shown only under the condition of fault occurrence, so that the condition of early shutdown or overtime work is shown. The invention adopts a large amount of data in the historical state of the equipment to establish a prediction model, and the fault condition of the equipment is identified by the difference value between the measured value of the equipment and the predicted value of the model, so as to predict the equipment fault on the basis of the difference value.
In this embodiment, the work of establishing the model is mainly completed by the model construction sub-module of the machine learning module. The fault diagnosis model constructed by the module is a set consisting of models and can be expressed as S ═ model1，model2，…，modeliThe elements in the set are different types of regression prediction models, and one type of model will predict the fault condition of a specific device. The machine learning module analyzes the received data block DataBloc { X, T }, extracts a characteristic vector set matrix { X } and a data type { T }, and extracts the characteristic vector set matrix { X } and the data type { T }, and the data type { T } is determined according to the data type of the data block{ T } classifying the feature vector set matrix { X } into different categories, and then allocating the data of the different categories to corresponding regression prediction models in the model set.
In this embodiment, taking a single constructed model as an example, the constructed model is a linear model, and the invention improves the linear model, specifically, a lasso-based linear regression model. Because the normal operation time of the equipment is predicted according to the sewage treatment equipment and the environmental data, both the dependent variable and the independent variable are continuous real values, regression analysis needs to be carried out on the data, and the linear model is simple in form, easy to model and strong in interpretability, an improved linear model, namely a lasso-based linear regression model, is supposed to be adopted as a prediction model of the equipment fault.
The lasso-based linear regression model can be expressed as:
wherein, YiResponse variable for the ith test sample, specifically predicted working time, X, of the feature vector concentration deviceijThe ith test sample corresponding to the jth second regression parameter is specifically real-time data of equipment and environment in the data feature vector;
estimated values of the first regression parameters representing the lasso linear regression model βjAn estimate of a jth second regression parameter representing a lasso linear regression model;iis the error variable for the ith training sample,i∈N(0,σ2) (ii) a i represents the serial number of a test sample, namely the serial number of real-time data of equipment and environment in the data characteristic vector to be predicted, and j represents the serial number of a model regression parameter; p represents the number of model regression parameters.
The parameter estimation of the lasso method is:
wherein, yiThe response variable of the ith training sample is specifically the working time of equipment in the data characteristic vector in the training process; x is the number ofijα is the ith training sample corresponding to the jth second regression parameter, specifically the real-time data of the equipment and environment in the data feature vectoriβ first regression parameter of lasso linear regression model for ith training samplejThe j-th second regression parameter in the lasso linear regression model is represented, t is a constraint parameter, i represents the serial number of the training sample, and j represents the serial number of the model regression parameter; t denotes a constraint parameter.
In order to obtain a regression parameter set satisfying lasso constraint, a mean square error loss function based on lasso is used as an objective function, a large amount of equipment data are used for training a model to obtain a solution satisfying the minimum value of the objective function, and a model regression parameter is obtained, wherein the loss function is expressed as:
β in the above equation is a second regression parameter vector of length n that does not include coefficients of the intercept term, and θ is an input vector of length n +1 that includes coefficients of the intercept term θ0M is the number of training samples, n is the number of features, | | | β | | purple1L representing parameter β1Norm, λ, is the constraint term of the input vector.
In this embodiment, to find the minimum value of the loss function, a coordinate descent method is used. Although the loss function is a convex function and has a global optimal solution, the regular term is l1 norm, and the absolute value is not derivable, so that the method is not feasible by adopting a traditional machine learning optimization method such as gradient descent and a least square method, and a common method for solving the lasso convex optimal solution comprises a coordinate descent method and a minimum angle regression method, and the coordinate descent method is supposed to be adopted for solving the problem that the speed is higher, the influence degree on sample noise is smaller than that of the minimum angle regression method, the robustness is better, and the method is suitable for a complex data environment in an industrial process.
In this embodiment, for the linear regression model based on lasso constraint, the selection of the parameter t is very important, if the parameter t is too large, the effects of preventing model over-fitting and variable selection are not achieved, and if the parameter t is too small, the obtained solution is too sparse, which may result in model under-fitting. Therefore, in order to select a more appropriate constraint parameter t, the model is trained by using a rolling cross validation method, and an appropriate parameter value is selected in the training process, wherein the training process comprises the following steps:
dividing the T training data into three parts of equal size T1, T2 and T3;
t1 partial data is selected as constraint parameter T with initial value
βiThe model regression parameters obtained by the least square method. The regression parameter represents a regression parameter solution of a common linear model without lasso constraint, and in order to achieve the effect of constraining the parameter, the constraint parameter t is decreased by a certain step length and is used as the constraint parameter of a new model to train the new model, so that evaluation models with different parameters t are obtained;
the T2 partial data is used for predicting and evaluating the evaluation model obtained by the T1 partial data, and prediction is carried out once when a sample observation value is sequentially added, so that the square sum SE of prediction errors of the T1 evaluation models is obtained;
in the formula of
Predicted value, yiI is the actual value and the serial number of the sample.
And the T3 partial data is used for obtaining a final parameter T, the parameter of the model with the lowest SE and the parameter weighted and averaged according to the SE are taken to train the training data consisting of the T1 and the T2, the obtained model is subjected to prediction evaluation by using the T3, and the parameter with the lowest SE is the final constraint parameter T. The method aims to compare the parameter with the minimum SE with the parameter weighted and averaged according to the SE, and the process shows that the parameter with the minimum SE is generated by using a certain step length and has the problem of step length precision, the average condition of a plurality of parameters can be integrated to a certain extent by using the parameter which is taken as the weight value according to the size of the SE, so that the value of the parameter is possibly more accurate, and the parameter with the minimum SE is selected as the final parameter according to the evaluation result of T3.
And the obtained final parameter T is used for establishing an equipment prediction model used in a real environment, the parameter is used as a constraint parameter, the coordinate descent method is used for gradually iterating according to an objective function, the model is trained on all training data with the quantity of T, a regression parameter set of the model is obtained, and then the final prediction model is obtained.
d. And c, collecting real-time data of the sewage treatment equipment and the sewage treatment environment from the production environment, performing the data processing part in the step b, taking the obtained feature vector set as the input of the trained model, and predicting the running time of the equipment as the judgment basis for diagnosing the fault.
e. And calculating the residual absolute value of the actual running time and the predicted running time of the equipment in real time, comparing the residual absolute value with a threshold, and outputting fault early warning information if the residual absolute value is greater than the threshold.
In this embodiment, a large number of training samples collected in the historical working environment and state of the sewage treatment equipment are used for training, so that in the actual sewage treatment process, when the equipment works, a prediction model does not need to be obtained through real-time training, the existing prediction model is constructed according to the historical condition of the equipment, and the real-time running state of the equipment can be accurately tracked. The sewage treatment process has stage and time lag, and conditions such as external climate and industrial factors change, so that the start-stop process and the running time of the process equipment treated each time have certain fluctuation, but the working time has larger fluctuation during fault, when the sewage treatment equipment possibly has fault, the equipment running time predicted by the equipment prediction model does not accord with the actual running time of the equipment, a difference value is generated between the equipment running time and the actual running time, and the deviation represents the fault condition to a certain extent.
In this embodiment, the basis for predicting the occurrence of the diagnostic equipment failure is an equipment failure threshold, and the threshold is obtained based on historical failure information of the equipment. In the system, the total data storage server stores a large amount of sewage industrial data, including long-time equipment index data, factory building environment data, equipment working logs and the like. And obtaining the working time of each equipment failure and the time to work under the normal condition according to the data of the starting and stopping time, the failure times and the like of the equipment working history.
The absolute value of the residual error between the normal working time and the fault working time is taken as a data variable to be analyzed, so that the obtained data is distributed more intensively, and the data is easier to process. For the data of the working time residual absolute value under the fault condition, the problem is that if the current equipment running time has a certain deviation from the prediction time, the degree of the deviation is determined to be that the equipment has a fault, and the problem that the deviation is converted into an abstract problem and is an abnormal value detection problem in data analysis can be considered. For the fault diagnosis problem of the sewage treatment equipment, namely, the absolute value of the actual working time and the predicted working time, the absolute value of the residual error under the fault condition can be judged when the absolute value is the absolute value, and the abnormal value under the fault condition, namely, the normal working time is the absolute value, but the fault diagnosis problem is not taken as fault treatment.
Common abnormal value monitoring methods comprise a statistical empirical method, a boxplot model and the like, wherein the statistical empirical method generally has a good effect when data distribution obeys normal distribution, but the distribution of fault data in actual analysis does not obey the standard normal distribution and has certain deviation, so that the statistical empirical method is not adopted. The box line graph is identified by calculation according to actual data, and does not make restrictive assumption on the data (such as obeying normal distribution), so that the box line graph can truly represent the original appearance of the data distribution. The judgment standard is based on the quartile and the quartile difference, the quartile has certain robustness, is only relevant according to the distribution condition of the data, is objective in identifying abnormal values, and has certain wide applicability, so that the system monitors the working time of faults by using a box line graph model.
It will be appreciated that certain features of the fault diagnosis system and method of the present invention may be mutually incorporated and are not to be considered an exhaustive list of the invention.
Those skilled in the art will appreciate that all or part of the steps in the methods of the above embodiments may be implemented by associated hardware instructed by a program, which may be stored in a computer-readable storage medium, and the storage medium may include: ROM, RAM, magnetic or optical disks, and the like.
The above-mentioned embodiments, which further illustrate the objects, technical solutions and advantages of the present invention, should be understood that the above-mentioned embodiments are only preferred embodiments of the present invention, and should not be construed as limiting the present invention, and any modifications, equivalents, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (8)
1. A fault diagnosis system of sewage treatment equipment based on L asso regression is characterized in that the fault diagnosis system comprises a sewage treatment subsystem, a plurality of supervision clients and a plurality of data acquisition devices arranged in sewage treatment equipment and a sewage treatment pool in a sewage plant, wherein the sewage treatment subsystem comprises a central console, a communication server, a total data storage server and a data acquisition server;
the central console comprises a control module, a data interface module, a data classification module, a machine learning module and a data communication module;
the control module is used for controlling and coordinating the work flow and data flow among the modules, carrying out data transmission with each server through the concentrator, and issuing real-time information or early warning information of the sewage treatment equipment according to the collected data of each server;
the data interface module is used for receiving a data collection instruction of the control module and accessing a total data storage server according to the instruction to obtain corresponding data;
the data classification module is used for classifying the received sewage equipment data and sewage treatment environment data and taking the classified data as input data of the machine learning module;
the machine learning module is used for constructing a sewage treatment equipment fault prediction model based on L asso linear regression and predicting sewage treatment equipment fault information;
the machine learning module includes a plurality of sub-modules,
the model construction submodule is used for constructing a sewage treatment equipment fault prediction model based on L asso linear regression;
a model training submodule: training a fault prediction model of the sewage treatment equipment by using historical data of the sewage treatment equipment and the sewage treatment environment as training set data, and acquiring parameters of the fault prediction model;
the model prediction submodule predicts the running time of the equipment by utilizing a trained sewage treatment equipment fault prediction model based on L asso linear regression;
a run-time residual calculation submodule: calculating residual errors of historical fault equipment running time and normal equipment running time to obtain a residual error sequence and residual error absolute values of actual running time and predicted running time of real-time computing equipment;
a threshold construction submodule: calculating a first four quantile point Q1, a median Q2 and a third four quantile point Q3 of the residual sequence, wherein the IQR is 1.5 (Q3-Q1), the Q1-IQR is a lower inner limit, and the lower inner limit is a fault threshold;
and a fault information judgment submodule: comparing the absolute value of the residual error with a fault threshold value, and outputting fault early warning information if the absolute value of the residual error is greater than the fault threshold value;
the data communication module is used for receiving equipment failure prediction information of the machine learning module and position information of the sewage treatment equipment and real-time data of the sewage treatment equipment and the sewage treatment environment acquired by the data interface module, and packaging and transmitting the information to the communication server;
the total data storage server is used for storing historical water quality data, equipment running state data and sewage treatment subsystem running data and sending the historical water quality data, the equipment running state data and the sewage treatment subsystem running data to the central console through the data interface module;
the communication server is responsible for data communication between a data communication module of the central control console and each supervision client;
the data acquisition server is used for acquiring real-time data of the sewage treatment equipment and the sewage treatment environment acquired by the data acquisition devices.
2. The L asso regression-based sewage treatment equipment fault diagnosis system of claim 1, wherein the issuing of real-time information or early warning information according to the collected data of each server comprises the machine learning module packaging the sewage treatment equipment fault information predicted by the machine learning module, including a prediction result, a model identifier and time sequence information, and transmitting the prediction result, the model identifier and the time sequence information to a central control platform, the central control platform obtaining corresponding equipment information according to the predicted sewage treatment equipment fault information, including the position of a monitoring point where the equipment is located, current equipment state data and environmental data, and packaging the prediction information together with the equipment into early warning information, and transmitting the early warning information to a data communication module, the data communication module transmitting the early warning information to a communication server, and the communication server transmitting the early warning information to a supervisor client.
3. A fault diagnosis method for sewage treatment equipment based on L asso regression is characterized by comprising the following steps:
a. acquiring a large amount of data of sewage treatment equipment and sewage treatment environment from a plurality of data acquisition devices of the sewage treatment equipment and a sewage treatment pool in a sewage plant, wherein the data comprises the data of the sewage treatment equipment and the data of the sewage treatment environment;
b. performing data preprocessing on the equipment data according to the type of the equipment, namely standardizing the characteristics of each dimension of the data according to the distribution condition of the characteristics of the type of the equipment to obtain a characteristic vector set as training set data;
c. constructing a fault prediction model of the sewage treatment equipment based on L asso linear regression, training the model by using training set data, and acquiring regression parameters of the model;
d. acquiring real-time data of the sewage treatment equipment and the sewage treatment environment from a plurality of data acquisition devices of the sewage treatment equipment and the sewage treatment pool in a sewage plant, performing a data preprocessing process consistent with the step b, and predicting the operation time of the equipment by taking the obtained feature vector set as the input of a trained model;
e. and calculating the residual absolute value of the actual running time and the predicted running time of the equipment, comparing the residual absolute value with a fault threshold value, and outputting fault early warning information if the residual absolute value is greater than the fault threshold value.
4. The fault diagnosis method for sewage treatment equipment based on L asso regression is characterized in that the fault prediction model for sewage treatment equipment based on L asso linear regression in step c comprises:
wherein, YiResponse variable for the ith test sample, specifically predicted working time, X, of the feature vector concentration deviceijThe ith test sample corresponding to the jth second regression parameter is specifically real-time data of equipment and environment in the data feature vector;
an estimate of a first regression parameter representing a lasso linear regression model;
an estimate of a jth second regression parameter representing a lasso linear regression model;iis the error variable for the ith training sample,i∈N(0，σ2) (ii) a i represents the serial number of a test sample, namely the serial number of real-time data of equipment and environment in the data characteristic vector to be predicted, and j represents the serial number of a model regression parameter; p represents the number of model regression parameters.
5. The method for diagnosing the faults of the sewage treatment equipment based on the L asso regression is characterized in that the first regression parameters in the step c and the estimated values of the second regression parameters are obtained in a mode of:
wherein, yiThe response variable of the ith training sample is specifically the working time of equipment in the data characteristic vector in the training process; x is the number ofijα is the ith training sample corresponding to the jth second regression parameter, specifically the real-time data of the equipment and environment in the data feature vectoriβ first regression parameter of lasso linear regression model for ith training samplejThe j-th second regression parameter in the lasso linear regression model is represented, t is a constraint parameter, i represents the serial number of the training sample, and j represents the serial number of the model regression parameter; t denotes a constraint parameter.
6. The L asso regression-based sewage treatment equipment fault diagnosis method according to claim 5, wherein the constraint parameter t is calculated by using a rolling cross validation method, and the establishment process comprises the following steps:
dividing the training data with the number T into three parts of equal size T1, T2 and T3;
t1 partial data is selected as constraint parameter T with initial value
βiFor the model regression parameters obtained by the least square method, decreasing the constraint parameter t by a certain step length, and training the new model as the constraint parameter of the new model to obtain evaluation models with different parameters t;
the T2 partial data is used for predicting an evaluation model obtained by the T1 partial evaluation, and prediction is carried out once every time a sample observation value is sequentially added, so that the square sum SE of prediction errors in all the evaluation models of the T1 partial data is obtained;
the T3 partial data is used for obtaining final parameters, parameters of an evaluation model with the lowest SE are taken, training data consisting of T1 and T2 is trained according to the parameters of SE weighted average, the obtained model is used for prediction evaluation by using T3 partial data, and the parameters corresponding to the minimum SE are used as the final parameters.
7. The method of claim 4, wherein a lasso loss function is used as the objective function, the prediction model is trained using a large amount of data to solve the minimum value of the objective function, and the estimated values of the first regression parameters and the estimated values of the second regression parameters of the prediction model are obtained to obtain the regression parameter set satisfying the lasso constraint, wherein the loss function is expressed as:
β in the above equation is a second regression parameter vector of length n that does not include coefficients of the intercept term, and θ is an input vector of length n +1 that includes coefficients of the intercept term θ0M is the number of training samples, n is the number of features, | | | β | | purple1L representing parameter β1Norm, λ, is the constraint term of the input vector.
8. The fault diagnosis method for sewage treatment equipment based on L asso regression is characterized in that the obtaining process of the fault threshold value in step e comprises the following steps:
calculating residual absolute values of historical fault equipment running time and normal equipment running time to obtain a residual absolute value sequence;
establishing a box-type graph statistical model, calculating a first four-quantile point Q1, a median Q2 and a third four-quantile point Q3 of a residual absolute value sequence, and making the IQR (iQR) 1.5 (Q3-Q1) and Q1-IQR as a lower inner limit; the lower inner limit is taken as the failure threshold.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910635096.1A CN110320892B (en) | 2019-07-15 | 2019-07-15 | Sewage treatment equipment fault diagnosis system and method based on L asso regression |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910635096.1A CN110320892B (en) | 2019-07-15 | 2019-07-15 | Sewage treatment equipment fault diagnosis system and method based on L asso regression |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110320892A CN110320892A (en) | 2019-10-11 |
| CN110320892B true CN110320892B (en) | 2020-07-14 |
Family
ID=68123565
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910635096.1A Active CN110320892B (en) | 2019-07-15 | 2019-07-15 | Sewage treatment equipment fault diagnosis system and method based on L asso regression |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110320892B (en) |
Families Citing this family (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110705133B (en) * | 2019-11-06 | 2022-09-23 | 中国联合网络通信集团有限公司 | Predictive maintenance method and predictive maintenance equipment |
| CN112818679A (en) * | 2019-11-15 | 2021-05-18 | 阿里巴巴集团控股有限公司 | Event type determination method and device and electronic equipment |
| CN110954348A (en) * | 2019-11-22 | 2020-04-03 | 东莞理工学院 | Factory sewage treatment fault diagnosis device |
| CN111027615B (en) * | 2019-12-05 | 2022-06-10 | 江苏方天电力技术有限公司 | Middleware fault early warning method and system based on machine learning |
| CN111401582B (en) * | 2020-03-12 | 2020-11-10 | 中交疏浚技术装备国家工程研究中心有限公司 | Abnormity identification method and monitoring platform for domestic sewage treatment facility |
| CN111209708B (en) * | 2020-04-22 | 2020-08-14 | 西南交通大学 | Machine learning-based pile-soil interaction prediction analysis method |
| CN112001440A (en) * | 2020-08-20 | 2020-11-27 | 苏州鸿哲智能科技有限公司 | Fault diagnosis logic algorithm and system |
| CN113176530B (en) * | 2020-08-25 | 2023-05-05 | 北京合众伟奇科技股份有限公司 | On-line electricity meter batch fault diagnosis method based on operation characteristics of dismantling meter |
| CN111967684B (en) * | 2020-08-27 | 2024-02-02 | 北京合众伟奇科技股份有限公司 | Metering asset active distribution method based on big data analysis |
| CN112364069A (en) * | 2020-09-14 | 2021-02-12 | 光大环境科技（中国）有限公司 | Thermocouple fault early warning method and system based on time sequence and storage medium |
| CN112305184B (en) * | 2020-10-12 | 2023-10-13 | 桂林理工大学 | Sewage treatment multi-fault diagnosis system and method |
| CN112540580A (en) * | 2020-10-30 | 2021-03-23 | 红云红河烟草(集团)有限责任公司 | Method for monitoring equipment state of wire making and feeding system based on historical production data |
| CN112363476A (en) * | 2020-11-16 | 2021-02-12 | 桂林理工大学 | Underground drainage monitoring system and method based on positioning technology |
| CN113393070A (en) * | 2020-11-18 | 2021-09-14 | 广州市八达工程有限公司 | Scheduling control method, device, equipment and storage medium for sewage treatment |
| CN112380206B (en) * | 2020-11-25 | 2022-07-19 | 浙江中控信息产业股份有限公司 | Diagnosis and repair method of traffic time sequence data |
| CN112667710A (en) * | 2020-12-24 | 2021-04-16 | 深圳市英威腾电气股份有限公司 | Inverter overheating early warning method and device, computer equipment and storage medium |
| CN112508457A (en) * | 2020-12-25 | 2021-03-16 | 树根互联技术有限公司 | Data processing method and device, industrial equipment and storage medium |
| CN113159597B (en) * | 2021-04-29 | 2022-08-16 | 浙江陀曼云计算有限公司 | Machine tool production interval calculation method and system based on time sequence data fluctuation |
| CN115129019A (en) * | 2022-08-31 | 2022-09-30 | 合肥中科迪宏自动化有限公司 | Training method of production line fault analysis model and production line fault analysis method |
| CN115578015B (en) * | 2022-12-07 | 2023-05-23 | 华南理工大学 | Sewage treatment whole process supervision method, system and storage medium based on Internet of things |
| CN116110516B (en) * | 2023-04-14 | 2023-07-21 | 青岛山青华通环境科技有限公司 | Method and device for identifying abnormal working conditions in sewage treatment process |
Family Cites Families (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104091035B (en) * | 2014-07-30 | 2017-02-08 | 中国科学院空间应用工程与技术中心 | Health monitoring method for effective loads of space station based on data-driven algorithm |
| CN104616124B (en) * | 2015-03-06 | 2018-08-17 | 重庆大学 | A kind of real-time release method and system of thermal power plant safety and economic operation state |
| US10935962B2 (en) * | 2015-11-30 | 2021-03-02 | National Cheng Kung University | System and method for identifying root causes of yield loss |
| CN105975443B (en) * | 2016-05-04 | 2019-02-15 | 西南大学 | Network anomalous behaviors detection method and system based on Lasso |
| CN106769040B (en) * | 2016-12-14 | 2019-02-26 | 上海工程技术大学 | A kind of method of the sparse reconstruct of bearing vibration signal |
| CN107065799A (en) * | 2017-04-19 | 2017-08-18 | 刘飞 | Long-distance monitoring method and its system for on-line monitoring sewage disposal system |
| CN107436597B (en) * | 2017-07-17 | 2019-10-18 | 华南理工大学 | A kind of chemical process fault detection method based on sparse filtering and logistic regression |
| CN207335786U (en) * | 2017-11-03 | 2018-05-08 | 浙江华成市政公用工程有限公司 | A kind of domestic sewage in rural areas remote intelligent operational system |
| CN109459993B (en) * | 2018-12-06 | 2021-06-25 | 湖南师范大学 | Online adaptive fault monitoring and diagnosing method for process industrial process |
-
2019
- 2019-07-15 CN CN201910635096.1A patent/CN110320892B/en active Active
Non-Patent Citations (3)
| Title |
|---|
| C. Cantelmo;L. Piroddi.Adaptive model selection for polynomial NARX models.《IET Control Theory & Applications ( Volume: 4 , Issue: 12 , December 2010 )》.2010, * |
| Junchi Yan;Minglei Tong.Weighted sparse coding residual minimization for visual tracking.《2011 Visual Communications and Image Processing (VCIP)》.2011, * |
| Rui Li;Zhenyu Liu;Jianrong Tan.Human motion segmentation using collaborative representations of 3D skeletal sequences.《IET Computer Vision ( Volume: 12 , Issue: 4 , 6 2018 )》.2018, * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110320892A (en) | 2019-10-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110320892B (en) | Sewage treatment equipment fault diagnosis system and method based on L asso regression | |
| CN109240244B (en) | Data-driven equipment running state health degree analysis method and system | |
| CN104992270B (en) | Power transmission and transformation equipment state overhauling aid decision-making system and method | |
| CN111160685A (en) | Maintenance decision method based on equipment comprehensive health condition analysis and management | |
| CN100470416C (en) | Power plant thermal equipment intelligent state diagnosing and analyzing system | |
| CN115034578A (en) | Intelligent management construction method and system for hydraulic metal structure equipment based on digital twinning | |
| CN107256449B (en) | State evaluation and assessment method for intelligent substation relay protection device | |
| CN113036913B (en) | Method and device for monitoring state of comprehensive energy equipment | |
| CN103559648A (en) | Grid equipment state inspection and evaluation training system | |
| CN107544457A (en) | Refinery plant running cycle expert decision system and method based on fail-safe analysis | |
| CN117176560B (en) | Monitoring equipment supervision system and method based on Internet of things | |
| CN103309342A (en) | Safety verification scheme aiming at industrial control system | |
| CN112415947A (en) | CNC machine tool data acquisition and management method and system based on DTU equipment | |
| CN106570567A (en) | Main network maintenance multi-constraint multi-target evaluation expert system and optimization method | |
| CN115063100A (en) | Data analysis-based implementation cost management and control system for road construction project | |
| CN116383636A (en) | Coal mill fault early warning method based on PCA and LSTM fusion algorithm | |
| CN110033181B (en) | Power generation equipment state evaluation method based on self-encoder | |
| CN112580858A (en) | Equipment parameter prediction analysis method and system | |
| CN117032120A (en) | Integrated intelligent cloud control system and control method for air compression station | |
| CN116644929A (en) | Intelligent management method for operation and maintenance work orders of power distribution network equipment based on Internet of things perception | |
| CN110533213A (en) | Transmission line of electricity defect Risk Modeling and its prediction technique based on support vector machines | |
| CN115407712A (en) | Intelligent maintenance system for hydraulic station of steel mill and working process | |
| CN115907033A (en) | Method and system for predicting hourly power consumption based on machine learning algorithm | |
| CN115936663A (en) | Maintenance method and device for power system | |
| CN111061191B (en) | Distributed oil-gas storage tank remote operation and maintenance method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |